Biochemical Characterization of Drosophila Receptor Tyrosine Phosphatases
Thesis by
Bruce Seymour Burkemper
In Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
California Institute of Technology Pasadena, California
2003
(Defended February 4th, 2003)
Acknowledgments
I’m grateful for my good fortune in having had the privilege to be a member of the Caltech community. It has been immensely gratifying to be at an institution so renowned for its pursuit of excellence. Working among people who have not only a love of science, but minds keen enough to deal with its complexity has been a humbling experience. Of all the people I have worked with, two stand out as most embodying the spirit of Caltech, and they are the ones I would most like to thank. The first is my thesis advisor, Kai Zinn.
Kai is the most erudite person I know, with an astounding memory and an uncanny ability for making astute observations. He also happens to be a skilled raconteur, so being in his company has always been interesting. I would very much like to thank him for his support and guidance, without which this thesis would not have been possible. I would also like to thank Peter Snow for applying his expertise in biochemistry to my projects.
Peter lives for science, and it was very inspiring to work with someone with such dedication. The contributions made by him are the foundations upon which much of my thesis work is built.
Having a benevolent thesis committee has enriched my experience as a graduate student. I thank the members of my committee, Marianne Bronner-Fraser, Bruce Hay and Mary Kennedy, for taking an interest in my projects and giving me advice to help my work come to fruition.
Finally, I would like to acknowledge the generosity of the following people for their contributions: Corey Goodman (UC Berkeley) for Robo reagents, Paul Garrity (MIT) for the DA3 construct, Neil Kreuger (Harvard) for the DLAR construct, Gary Hathaway (Caltech PPMAL facility) for sequencing and mass spectrometry, Jody Franke and Dan Kiehart (Duke) for myosin II-tail-coated beads, and Liquin Luo (Stanford) for mushroom body reagents.
Abstract
Two classes of enzymes are responsible for modulation of intracellular
phosphotyrosine levels, namely protein tyrosine kinases (PTKs) and protein tyrosine phosphatases (PTPs). Together these enzymes maintain the appropriate balance of phosphoproteins required for a variety of developmental processes including axon pathfinding. In Drosophila, five receptor-like protein tyrosine phosphatases (RPTPs) regulate axon pathfinding, but little is known about their downstream signaling pathways or the means by which their enzymatic activity is regulated.
Chapter 2 of this thesis deals with experiments to test whether dimerization
regulates the activity of these enzymes. Crystallographic data indicates that some RPTPs form dimers in which each monomer is precluded from binding substrate due to the insertion of a helix-turn-helix segment of the opposing monomer into the active site. I introduced “tagged” RPTP constructs into Drosophila S2 tissue culture cells and tested for dimer formation using immunoprecipitation and Western blotting. I did not detect stable dimers, however. This may suggest that dimer formation requires other protein components (such as the putative RPTP ligands) that are not expressed in S2 cells.
In Chapter 3 I investigated the possibility that Roundabout (Robo), a receptor mediating axonal repulsion from the embryonic midline, is a substrate for RPTPs DPTP69D and/or DPTP10D. Previous genetic studies implicate these RPTPs in
participating in the Robo signaling pathway. Experiments detailed here show that Robo can be phosphorylated on tyrosine residues in S2 cells, characteristic of an RPTP
substrate. However, Robo did not co-immunoprecipitate with “substrate trap” mutants of either of these RPTPs, possibly because their interaction is dependent on co-factors not present in the cell culture system.
Chapter 4 is a characterization of DPTP69D-associated proteins purified from embryos expressing a substrate trap version of DPTP69D. We identified one of the associated proteins as non-muscle myosin II heavy chain (nmm II hc). Proper regulation of nmm II hc is essential for axon patterning in mushroom bodies (MBs). I found that expression of the DPTP69D trap in MBs results in an axon retraction phenotype similar
to that seen when nmm II hc activity is elevated, suggesting that this protein may be a target for DPTP69D activity.
Table of Contents
Acknowledgments ii
Abstract iii
Chapter 1 Structure and function of RPTPs A 1
Chapter 2 Analysis of RPTP dimerization potential in Drosophila S2 cells B 1
Chapter 3 Biochemical analysis of potential interactions between RPTPs C 1 and Robo receptors
Chapter 4 Studies of proteins purifying with a “substrate trap” version of D 1 DPTP69D
References E 1
Chapter 1
Structure and function of receptor tyrosine phosphatases
Introduction
The evolutionary transition from unicellular to multicellular organisms was predicated on the development of a mechanism for cell-cell signaling. Transmembrane signaling events underlie a myriad of developmental processes including morphogenesis, pattern
formation and differentiation. The adult multicellular organism is also dependent on cell- cell signaling for many aspects of normal cell maintenance and functioning. A variety of extracellular stimuli such as cytokines, growth factors and hormones activate receptor protein tyrosine kinases (PTKs) leading to an increase in intracellular phosphorylation levels. PTKs are also activated through various types of cell-cell, cell-matrix interactions.
The resulting phosphorylation of select intracellular target proteins triggers downstream signaling pathways that effect diverse responses through generation of second
messengers and activation of other proteins, such as Ser/Thr kinases and G-proteins.
Tyrosine phosphorylation is a viable means of manifesting cell-cell signaling only if the phosphorylation is reversible. In general, once a given signal has caused it to increase, the level of intracellular phosphotyrosine must return to a baseline in order for the cell to respond to subsequent signals. Equally important as PTKs, then, are the protein tyrosine phosphatases (PTPs) that subserve this function. Similar to PTKs, the importance of these enzymes is reflected in their diversity. PTPs can be grouped into two structurally distinct groups: receptor-like proteins that span the membrane and soluble cytosolic enzymes.
Receptor-like protein tyrosine phosphatases (RPTPs) have the potential to transduce extracellular signals into changes in phosphotyrosine levels via ligand binding-induced activation of their catalytic domain. Instead of simply reversing phosphorylation
mediated be kinases, RPTPs may play an active role in regulating the cellular response to outside signals.
In the following sections I will review our current knowledge of the RPTPs. (1) Certain residues are highly conserved in the RPTP catalytic domain and known to be essential for catalytic activity. Our current understanding of the basic enzymatic
mechanism for RPTPs is supported by a large body of empirical evidence. Elucidation of this mechanism has contributed to the development of mutant trap versions of these
enzymes that have been useful in identifying potential RPTP substrates. (2) RPTPs have well-defined preferences for substrates. It has been shown that protein tertiary structure is not as important as the primary sequence adjacent to the phosphotyrosine residue in determining the suitability of a given phosphopeptide as a substrate. (3) Interactions within and between RPTPs are an important means by which the activity of these enzymes is regulated. In particular, the “wedge” region plays a critical role in
extinguishing enzymatic activity. (4) With few exceptions, RPTPs have two catalytic domains, D1 and D2. Most activity resides in D1, but the low level of activity in D2 can be increased to levels comparable to D2 through mutation of just two residues.
1. Enzymatic mechanism
All “classical” tyrosine phosphatases have a conserved catalytic domain of approximately 240 residues characterized by the PTP signature motif
(I/V)HCXAGXXR(S/T)G. This motif contains invariant cysteine (Cys) and arganine (Arg) residues that are essential for catalytic activity. The role of the Arg is to assist in positioning the phosphotyrosine substrate (pTyr) such that its phosphorus atom is situated adjacent to the sulfur atom of the PTP motif’s catalytic Cys residue (Jia et al., 1995). This allows Cys to launch a nucleophilic attack on the phosphorus atom, the first step in the dephosphorylation reaction.
The PTP motif is situated near the center of the molecule and is surrounded by four loops that delineate the entrance to the active site: L1, L6, L13 (or WpD) and L17.
Binding of substrate results in a conformational change in the WpD loop from an open position into a catalytically competent closed position. This brings the side chain of an invariant, catalytically essential aspartic acid on the loop (Asp 181 in PTP1B) into position to act as a general acid. The closed WpD conformation is stabilized by a combination of hydrogen bonds and hydrophobic interactions.
Attack of the Cys nucleophile on phosphorus results in a pentavalent transition state.
In donating a proton to this transition state, Asp 181 helps to cleave the P-O bond between the phosphorus and the oxygen of the tyrosyl side chain in the substrate. The result is formation of a thiol-phosphate (phosphoenzyme) intermediate.
The final step in the dephosphorylation reaction is hydrolysis of the thiol-phosphate intermediate by an activated water molecule. A glycine residue in the PTP facilitates this step by positioning the nucleophilic water molecule adjacent to the phosphorus atom of the intermediate. Gln 262 participates in hydrogen bonding with the water molecule to give it the proper orientation. Next, the same Asp that earlier served as a proton donor now acts as a proton acceptor and abstracts a hydrogen from the water molecule. This converts it into an active nucleophile that attacks the phosphorus atom, thereby cleaving the cysteinyl-phosphorus bond and reconstituting the enzyme. A schematic of the PTP dephosphorylation reaction is shown in Figure 1.
Several lines of evidence support this mechanism of catalysis. Site directed mutagenesis of the catalytic Cys results in an inactive phosphatase in an assay with 32P- labeled substrates (Guan and Dixon, 1991). When the invariant Asp is mutagenized, there is a marked reduction in catalytic activity (Zhang et al., 1994b) and substrates remain bound to the enzyme (Flint et al., 1997). The latter finding is consistent with the proposed role of Asp in cleaving the pTyr P-O bond that would cause release of the
dephosphorylated substrate. We have exploited the substrate trapping properties
conferred by this mutation in a search for RPTP substrates (see Chapters 3 and 4). Finally, substitution of Gln 262 for Ala leads to accumulation of the thiol-phosphate intermediate, consistent with the role of this residue in catalyzing hydrolysis of the intermediate (Denu et al., 1996).
2. Substrate binding
Several criteria determine the suitability of a given protein as a substrate for PTPs.
The most important is that it is phosphorylated on one or more tyrosine residues. PTPs display a rigid specificity for pTyr-containing substrates, and proteins phosphorylated on other residues are unable to be dephosphorylated by these enzymes. This distinguishes the PTPs from Ser/Thr phosphatases and the dual-specificity phosphatases, the latter being able to dephosphorylate proteins phosphorylated on Ser, Thr and Tyr residues.
To determine the other criteria for PTP substrates, phosphorylated synthetic peptides are commonly used as model substrates. This approach is more practical than
using physiological substrates because few naturally occurring pTyr-containing substrates are readily available. Also, the broad specificity of kinases can be exploited to
phosphorylate a diverse collection of artificial peptide substrates in vitro. For example, a series of peptides differing by just one residue can be kinased and tested with a PTP to determine the relative importance of each residue in the peptide for the
dephosphorylation reaction.
Dimensions of the catalytic site cleft are the primary determinant of the molecular basis for PTP substrate specificity (Dunn et al., 1996). The nucleophilic Cys is situated at the base of this cleft, and there is a distance of 9 Aº from it to the cleft entrance. A
conserved KNRY sequence forms a phosphotyrosine recognition loop that contributes to formation of the cleft (Jia et al., 1995). The Tyr residue from this loop and other nonpolar residues from the WpD loop interact with the phenyl ring of the substrate peptide’s pTyr.
These interactions cause the pTyr residue to adopt a helical conformation that inserts into the cleft. The importance of the cleft’s depth was documented in experiments with a peptide of the general form (Glu)4-NH-(CH2)n-PO3. The peptide that was most efficiently dephosphorylated was the one where n=7, which corresponds exactly to the length of a tyrosine residue.
Studies of several PTPs have determined that pTyr site recognition depends not on higher orders of protein conformation but on the primary sequence surrounding pTyr, particularly amino acids immediately N- and C-terminal to this residue. For a particular pTyr-containing substrate, the rate of dephosphorylation by PTP can be easily measured with a continuous spectrophotometric assay due to the different absorption spectra of pTyr versus Tyr. Several PTPs have been examined with this assay to define precisely which substrate residues are important for efficient dephosphorylation.
In the case of PTP1 and Yersinia PTPase, four residues N-terminal and one residue C-terminal to pTyr were found to be key determinants of substrate suitability (Zhang et al., 1994a). This was determined by sequentially substituting Ala for each amino acid within a pTyr-containing peptide substrate derived from the EGF receptor (EGFR988-998: DADEpYLIPQQG), and measuring the kinetics of the dephosphorylation reaction.
The -1 position immediately N-terminal to pTyr was found to be particularly important, as a substitution of the Glu residue here for Ala resulted in a 126-fold decrease in enzyme
activity on this peptide compared to wild-type peptide. Additional experiments with various length substrates determined that a minimum of six amino acids is required for optimal binding to these PTPs, including the pTyr residue.
In general, PTPs have a documented preference for substrates with acidic residues N-terminal to the pTyr due to the presence of basic residues on the PTP surface that interact with these residues. The presence of pTyr is necessary but not sufficient for high- affinity binding to PTPs, as a singular pTyr residue binds only weakly, and peptides without pTyr do not bind at all (Barford et al., 1998).
3. Modulation of activity
Regulation of intracellular phosphorylation levels is a fundamental mechanism used by cells to regulate a wide array of cellular functions including proliferation, differentiation and metabolism. In axon patterning, RPTP activity is required at specific points along the path of extending axons to keep them oriented towards their target. It is clear that the involvement of PTPs in these processes requires that their activity be specifically regulated. This is accomplished in some measure through proteins that target the enzymes to specific subcellular locations. Other mechanisms of regulation have yet to be discovered.
PTPs are efficient catalysts, and isolated catalytic domains exhibit constitutive activity. The pathogenicity of the Yersinia bacteria responsible for tuberculosis and bubonic plague is due to disruption of signal transduction pathways resulting from constitutive activity of the Yop family PTPs that it injects into macrophages. Thus, regulation of PTP activity may come in the form of inhibition of their activity.
Dimerization has been proposed as one means by which PTP activity may be inhibited. The first evidence supporting this theory comes from crystallographic studies of RPTPα. Like most phosphatases, RPTPα consists of two catalytic domains, D1 and D2, each with an active site. The crystal structure of the first catalytic domain indicates a region of each catalytic domain, denoted the “wedge,” inserts into the active site of the neighboring RPTPα molecule (Bilwes et al., 1996). These results suggest that dimer formation may represent a way of reversibly suspending phosphatase activity, as a
blocked active site would preclude substrate binding. In this scenario, RPTPs would be active only when they have been dissociated from dimers into monomers (Figure 2).
The wedge of RPTPαD1 is formed from a helix-turn-helix segment in the N- terminal segment of each monomer. The interhelical angle is ~80º, forming a wedge that inserts into the catalytic cleft of the opposing monomer. Hydrogen bonds and van der Waals interactions between the N-terminal wedge of one monomer and residues from the WpD loop of its dimer partner stabilize the WpD loop in the open conformation. This prevents the loop from moving into the catalytically competent closed position.
Additionally, the side chains of several N-terminal wedge residues interact with residues Tyr 262 and Asn 264 on the L1 loop of the opposite dimer. These are the equivalent of residues Tyr 46 and Asp 48 in PTP1B, two residues that have been shown to form critical interactions with the pTyr-containing substrate. This suggests dimerized RPTPαD1’s are incapable of binding substrate.
Taken together, the crystallographic data indicate that activity is proscribed in the dimeric form of RPTPαD1 on several levels. The catalytic site is physically occluded by the wedge, an Asp on the WpD loop essential for activity is kept away from the active site by interactions holding the loop in an open conformation, and residues that
participate in substrate binding are otherwise engaged in interactions with the L1 loop.
The sequences corresponding to the wedge show a high degree of conservation among other RPTP family members, suggesting that dimerization as a means of activity regulation could be a strategy common to many RPTPs.
Additional evidence for a model of dimer formation as a means of reversibly suspending RPTP activity comes from studies of an EGFR/CD45 chimera (Desai et al., 1993; Majeti et al., 1998). CD45 is an RPTP with no known ligand that is expressed on nucleated hematopoietic cells and is required for TCR signaling in response to
engagement of antigen receptor. To study the effects of dimerization on CD45 activity, a chimeric protein was made with the extracellular and transmembrane domains of EGFR fused to the intracellular domain of CD45. TCR signaling is normal in cells expressing the chimeric protein, but addition of EGF results in an abrupt loss of signaling. Signaling is restored in the presence of EGF when the chimera is coexpressed with a gene encoding only the extracellular and transmembrane portions of EGFR. These results suggest that
induced dimerization extinguishes the catalytic activity of the chimeric protein. Under physiologic conditions, ligand may substitute for EGF in inducing dimerization of wild- type CD45. Consistent with this, the wedge domain of CD45 is highly conserved with that of RPTPα, suggesting it may form dimers in a manner similar to RPTPα.
Other studies have shown dimer interactions can be complex, with a domain from one RPTP inserting into the catalytic domain of a different RPTP to form cross-species heterodimers (Blanchetot and den Hertog, 2000; Gross et al., 2002). For example, the second catalytic domain (D2) of PTPδ has been shown to bind to D1 of RPTPσ, and this results in ~50% reduction in catalytic activity of the latter (Wallace et al., 1998).
Evidence also exists for formation of multimers, where strings of RPTPs are
interconnected via interactions between wedge and active site domains of neighboring molecules (Iversen et al., 2002).
Strong evidence exists for a model of negative regulation of activity through dimerization for the RPTPs cited above, but the crystal structures of some other RPTPs suggests this is not a universal strategy employed by all RPTP family members. For example, RPTPµ does exist as a dimer in the crystal structure, but the wedge domain of one subunit of the dimer is not inserted into the catalytic cleft of the dyad-related monomer. Consequently, the active site remains in an open, uninhibited conformation (Hoffmann et al., 1997). In the case of RPTP LAR there is no direct evidence for a dimer, but other types of intramolecular interactions are thought to occur. The crystal structure suggests D1 of one LAR molecule may interact with D2 of a second molecule (Nam et al., 1999).
4. D1 vs. D2
Most RPTPs have two tandem phosphatase domains denoted D1 and D2. Studies of human RPTP LAR indicate the tertiary structure of LAR D1 and D2 are very similar (Nam et al., 1999). However, as is the case with other RPTPs, a majority of the catalytic activity resides in the membrane-proximal D1, with membrane-distal D2 possessing little or no activity. This raises two questions that apply to all members of the RPTP family: 1) what accounts for the disparity in catalytic activity between D1 and D2, and 2) if D2 is
not catalytically active, what function(s) does it serve? The high degree of primary sequence conservation among RPTP D2 domains suggests there does exist an as yet undiscovered function.
The most obvious difference between LAR D1 and D2 is the presence of four additional residues in D2 located in the loop between helices α1' and α2'. Alignment with sequences of other RPTPs shows the extra residues are shared by many D2 domains, suggesting the longer loop resulting from these additional residues may have biological significance.
Overall similarity is observed in the active site of both domains, each comprised of a catalytic cleft surrounded by four loops. Closer inspection reveals a substitution of two amino acids with important roles in the catalytic reaction. In PTP1B, an Asp on the WpD loop is thought to act as a general acid when the loop assumes the catalytically competent closed conformation. By donating a proton to the tyrosyl oxygen, Asp 181 assists in the reaction by turning the soon-to-be dephosphorylated substrate into a favorable leaving group. In D2 of LAR, the residue equivalent to Asp 181 is replaced by Glu 1779. The second substitution is found in the pTyr recognition loop that has been shown to interact with the substrate pTyr. In PTP1B, Tyr 46 forms hydrogen bonds with the Ser residue immediately adjacent to the catalytically essential Cys. The LAR residue corresponding to Tyr 46 is Leu 1644, which does not participate in hydrogen bonding with Ser. The absence of this interaction causes Ser to shift slightly into the path of the catalytic cleft, thereby precluding potential substrates from gaining access.
Mutational analysis was performed to determine whether these amino acid substitutions were responsible for the lack of phosphatase activity in D2. Individual substitutions of Glu 1779 for Asp or Leu 1644 for Tyr resulted in detectable
dephosphorylation of a 32P-labeled synthetic peptide, but the level of activity was a small fraction of that of wild-type enzyme. When both substitutions were made together, a far greater impact was observed, with D2 now showing activity levels equivalent to that of D1. Conversely, when the Asp and Tyr residues of D1 are mutated, D1 catalytic activity is abolished. Thus, the difference in catalytic potential of these domains is accounted for in its entirety by only these two residues.
Amino acid substitutions that correspond to those made in LAR D2 also result in fully restored activity of the D2 domain of RPTPα , suggesting this is a common means for D2 inactivation (Lim et al., 1998). Although the D2 domain plays little or no role in catalysis, the high degree of sequence conservation among RPTP family members suggests they do have an important function. One possibility is that D2 is involved in regulating the substrate specificity of D1. Consistent with this, the N-terminal part of LAR D2 interacts with D1 via hydrogen bonds and van der Wall forces and forms a wall on one side of the D1 active site. Mutants missing the N-terminal part of D2 show altered substrate specificity, while no such effect is seen in mutants in which the C-terminal part is missing. A similar observation has been made in the case of CD45, where a 19 residue insertion between two β sheets of the D2 domain results in altered D1 specificity for synthetic peptides (Streuli et al., 1990).
D2 domains may also be important for substrate binding, with evidence in support of this role provided by studies with insulin receptor (IR), a physiological LAR substrate (Tsujikawa et al., 2001). LAR constructs missing either the D1 or D2 domain were coexpressed with IR in COS cells. Following addition of insulin to stimulate IR autophosphorylation, LAR was immunoprecipitated (IPd) with an antibody against its extracellular domain. IR co-IPd efficiently with constructs missing the D1 domain, but the amount co-IPing with constructs missing the D2 domain was negligible.
Figure 1. Schematic of the reaction mechanism catalyzed by PTP1B. (a) Formation of the cysteinyl-phosphate intermediate. (b) Hydrolysis of the cysteinyl-phosphate intermediate.
Source: Annual Review of Biophysics and Biomolecular Structure (1998) 27, 154
Figure 2. Stereo ribbon diagram of the RPTPαD1 dimer. The label AS placed near the catalytically essential Cys 433 emphasizes the active site of each monomer.
Source: Nature (1996) 382, 556
Chapter 2
Analysis of RPTP dimerization potential in Drosophila S2 cells
Abstract
The vital role of receptor-linked tyrosine phosphatases (RPTPs) in Drosophila axon pathfinding has been well documented, yet little is known about how the activity of these enzymes is regulated. Crystallographic data suggests activity may be regulated through dimerization, as the “wedge” domain of one RPTP appears to insert into the active site of its dimer partner, thereby precluding it from binding substrate. Genetic data from our lab provides corroborating evidence for a model of dimer mediated regulation of activity. To assay for dimer formation, we expressed affinity-tagged versions of the RPTPs in the Drosophila S2 cell line. No evidence for stable dimers was found under the conditions tested, possibly because dimer formation is ligand induced. Our experimental approach can be used to study the role of ligands in inducing RPTP dimerization once such ligands are identified.
INTRODUCTION
The process of axon guidance requires the maintenance of an appropriate balance of intracellular tyrosine phosphorylation levels, which is maintained by two classes of enzymes: protein tyrosine kinases (PTKs) and protein tyrosine phosphatases (PTPs).
Each class is comprised of a large family of multidomain proteins in both cytoplasmic and transmembrane receptor forms. Many receptor-linked tyrosine phosphatases (RPTPs) resemble cell adhesion molecules in that their extracellular domains consist of fibronectin type III (FN III) repeats and/or immunoglobulin-like (Ig) domains. This suggests RPTPs may be one means by which cell recognition events are coupled to changes in
cytoplasmic phosphorylation levels.
In Drosophila, five RPTPs have been identified that play a role in axon pathfinding.
DPTP69D, DPTP10D, DPTP99A, DLAR, and DPTP52F are exclusively expressed in CNS neurons during the period of axonogenesis, and flies with mutations in their genes show characteristic defects in axon guidance during embryonic development (Desai et al., 1996; Desai et al., 1997; Krueger et al., 1996; Schindelholz et al., 2001; Sun et al., 2000;
Sun et al., 2001). Four of the five RPTPs (DPTP69D, DPTP10D, DPTP99A and DLAR) are restricted to axons. An axon reaches its target by taking a series of steps past
intermediate targets, each successive target leading it closer to its final destination. The intermediate targets, called choice points, are places at which the axon makes a decision about the trajectory it will follow in the next segment of its journey. Modulation of RPTP activity at these points is thought to be required for changes in growth trajectories.
Successful navigation past a series of choice points to an end target thus depends on the coordinated regulation of these proteins. Despite extensive studies documenting the
pathfinding role played by the RPTPs, little is known about how their activity is regulated.
Crystallographic studies of mammalian RPTPs suggest that dimerization may regulate the activity of these enzymes (Jiang et al., 1999). Like most transmembrane phosphatases, RPTPα has two catalytic domains, D1 (membrane-proximal) and D2 (membrane-distal), each with an active site. The crystal structure for this protein indicates that a region of each D1 catalytic domain, denoted as the “wedge,” inserts into the active site of the neighboring RPTPα molecule (Bilwes et al., 1996). These results suggest that
dimer formation may represent a way of reversibly suspending phosphatase activity, as a blocked active site would preclude substrate binding. In this scenario, RPTPs would be active only when they have been dissociated from dimers into monomers. Signals present in the environment traversed by the axon during extension may serve to bring RPTP monomers together to temporarily extinguish their activity. Similarly, other signals may work to cause existing dimers to dissociate into monomers, thereby switching activity on.
A combination of these signals, at the appropriate time and place during axonogenesis, could play a central role in facilitating the navigation of axons to their targets.
Several considerations make regulation of Drosophila axonal RPTPs through dimerization an appealing model worth testing. First, four RPTPs (DPTP10D, DPTP69D, DPTP99A, DLAR) have sequences similar to the wedge found in RPTPα (Figure 1).
Second, previous work in the lab has revealed that a heteromultimer including DPTP10D and DPTP69D is capable of forming. Third, genetic data gathered in our lab indicates that DLAR suppresses the activity of DPTP99A (Desai et al., 1997). One way to explain this result is to invoke formation of a DLAR/DPTP99A heterodimer. Finally, a precedent for activity modulation through dimerization exists in RPTP counterparts, namely receptor tyrosine kinases (RTKs).
We tested the dimerization model by cotransfecting epitope-tagged RPTP expression constructs pairwise into Drosophila S2 cells, immunoprecipitating one and looking for evidence that the second co-IPd. This approach was used to assay for both hetero- and homodimer formation. Initial experiments with constructs containing only cytoplasmic RPTP domains showed no evidence for dimerization. A second round of experiments with full-length RPTPs yielded the same result. Relatively weak expression in the S2 cell culture system leaves open the possibility that negative results were due to an inability to detect low protein levels. More likely, formation of putative dimers depends on one or more co-factors not present in S2 cells.
MATERIALS AND METHODS Plasmid Construction
Epitope-tagged constructs, denoted psmyc and psha, were manufactured using the S2 expression vector pRmHA3 as starting material. The PCRd cytoplasmic domain of each RPTP was subcloned into these base constructs. The salient features of the constructs include a metallothionein promoter for induction of expression with copper sulfate. A translation start consensus sequence is located immediately upstream of the initiator methionine for efficient expression. Immediately downstream of the initiator ATG, a Src myristylation sequence is present to direct the expressed protein to the inner surface of the cell membrane. The multiple cloning site includes five restriction sites selected to accommodate subcloning of RPTPs. Following the MCS is a single myc, ha or rho epitope tag, and finally a stop codon. 3' PCR primers were designed to ensure that the coding sequence of each RPTP domain is in frame with the epitope tag.
The finished constructs were sequenced by the Caltech Sequencing Facility to confirm accurate amplification of the RPTP cytoplasmic domains. Full-length, untagged versions of DPTP10D, DPTP69D and DPTP99A were made by fellow lab member Sarah Fashena by subcloning cDNAs into pRmHa3. The cytoplasmic (in psmyc) and full-length (in pRmHA3) DLAR constructs were made by Neil Krueger at Harvard University.
Baculovirus constructs were made by cutting full-length RPTPs out of pRmHa3 and dropping them into the baculovirus vector pVL393.
DLAR mAb generation
A portion of the DLAR extracellular domain containing three FN III repeats was PCRd and subcloned into pVL393, a transfer vector containing a baculovirus promoter flanked by baculovirus DNA derived from the polyhedron gene. This was submitted to Peter Snow in the Protein Expression Facility, who manufactured the recombinant virus via homologous recombination. Dr. Snow purified the recombinant protein, which was then given to Susan Ou of the Monoclonal Antibody Facility. Monoclonal antibodies (mAbs) generated against the recombinant protein were initially screened with ELISA. Positive
clones were then tested for their ability to immunoprecipitate DLAR from cell lysates, and to recognize DLAR on Western blots. The mAbs 8C42F5 and 9D82B3 produced the best results and were used in combination for these experiments.
S2 cell transfection, induction, harvesting and lysis
Growth medium used for S2 cells consisted of Schneider’s medium supplemented with 10% heat inactivated fetal calf serum, penicillin (100U/ml), streptomycin (100 µg/ml) and amphotericin B (0.25 µg/ml). 107 cells were plated on 10 cm tissue culture plates and expanded overnight (25º C, atmospheric pressure). Cells were transiently transfected with the DNA of interest (10 µg) using the calcium phosphate method. 18 hrs later the cells were washed with PBS and resuspended in growth medium. 0.3 mM CuSO4 was added to induce the metallothionein promoter to drive expression. The cells were harvested 24 hrs later, washed in PBS and lysed in 450 µl of ice-cold lysis buffer (125 mM NaCl, 10 mM TrisCl pH 7.5, 0.2% Triton X-100, 600 µM PMSF, 2 mM Na3VO4, 25 mM NaF and 2 µg/ml APP). The lysate was spun briefly to pellet nuclei and insoluble membrane components.
Immunoprecipitations and Western blot analysis
Lysates were incubated with primary antibody and protein G+A agarose beads for 1 hr at room temperature. The agarose beads with bound immune complexes were spun down and washed with lysis buffer twice. The pellets were boiled for 3 minutes in SDS sample buffer. Samples were run on 9% polyacrylamide gels and transferred to a PVDF
membrane. Blots were blocked in 5% dry milk in TBST (25 mM TrisCl pH 7.4, 137 mM NaCl, 0.2% Tween 20) for 30 minutes at RT. The blots were incubated in primary antibody for 1 hr, washed with TBST, and incubated with an alkaline phosphatase- conjugated goat anti-mouse secondary. After 30 minutes washing the blots were developed to detect phosphatase activity.
RESULTS
Cytoplasmic RPTP domains insufficient for dimerization
We tested for dimerization by transiently cotransfecting two RPTPs into S2 cells and inducing their expression. One RPTP was IPd, and if the other co-IPd it was taken as evidence in support of dimerization. Because it contains the active site and wedge domain, the RPTP cytoplasmic domain was considered sufficient to mediate the presumptive dimerization. By excluding the extracellular and transmembrane domains, the possibility of two RPTPs sticking together nonspecifically and being misinterpreted as a dimer is reduced. The cytoplasmic domain of each RPTP was PCRd and subcloned into an
epitope-tagged S2 expression construct. A Src myristylation sequence was included in the construct to localize the expressed protein to the membrane. Other features of the
construct are depicted in Figure 2.
S2 expression constructs made for each of three different epitope tags (myc, rho and ha) were tested with a uniform insert to determine which tag yielded the best
immunoprecipitation results. The efficiency of IPs was roughly equivalent with the myc and rho epitopes, while the ha tag was considerably weaker. Unlike ha, myc and rho IPs were strong enough to provide results that could be interpreted reliably. RPTP
cytoplasmic domains were subcloned into the myc- and rho-tagged expression constructs.
Two constructs with dissimilar epitope tags were cotransfected in each experiment.
Following a myc IP, the rho-tagged RPTP is visualized on a Western blot with an anti- rho antibody. The experiment is performed in a complementary manner as well, by IPing with a rho antibody and looking for evidence of the myc-tagged RPTP with an anti-myc antibody.
One difficulty encountered early in the course of experiments was variable expression of the transfected RPTP constructs. For reasons unknown, some constructs expressed well when transfected individually, but not when coexpressed with a second RPTP. Under these conditions the expression levels would drop off to varying degrees.
The calcium phosphate method calls for 10 µg of the DNA being transfected, so in cotransfections 10 µg of each plasmid was used for a total of 20 µg. The two plasmids were introduced to the cells at the same time after mixing in a transfection solution.
Efficiency of transfection with calcium phosphate is roughly 10-20% for S2 cells. Certain measures were taken to increase the efficiency when dealing with a construct whose expression was inconsistent in cotransfections. This included increasing the amount of DNA (up to 20 µg), and lengthening the cell culture incubation time between addition of DNA and harvesting of cells.
Due to the variable expression of some constructs, it was important that our experimental design incorporate a positive control for protein expression. A result militating against dimerization is meaningful only when both RPTPs expressed well enough to be detected on a Western blot. It was also important to include a negative control to exclude antibody cross-reactivity as the reason for any observed co-IPs.
Results from a typical experiment are shown in Figure 3. In this experiment, two differently tagged versions of DLAR were cotransfected to look for evidence of DLAR homodimers. Two constructs expressing the DLAR cytoplasmic domain were transfected, one carrying a myc tag (DLARmyc) and the other a rho tag (DLARrho). Lysate from transfected cells was split in two; half was IPd with anti-myc and half with anti-rho. The myc IP was itself split in two, and the half probed with anti-myc shows a DLARmyc band, confirming expressed of DLARmyc. If DLARmyc complexed with any DLARrho, some of the latter should have come down with it. However, when the second half of this IP is probed with anti-rho, no DLARrho is visible. The negative result could be
accounted for if DLARrho failed to express, since that would preclude a
DLARrho/DLARmyc complex from forming. This possibility is eliminated when the second batch of lysate is IPd with anti-rho. The IP pellet is again split in two, and half is probed with anti-rho. The resulting DLARrho band confirms DLARrho did express. The other half of the pellet is probed with anti-myc. Despite the fact that DLARmyc and DLARrho were expressed alongside each other, the absence of a DLARmyc band here once again suggests that none of it formed a complex.
It is possible that our negative results are due to the fact that the cytoplasmic domain alone is insufficient to mediate dimerization. Perhaps the RPTP extracellular domain is necessary for the protein to assume the conformation required for forming dimers. This theory was tested by repeating the experiments, this time using constructs coding for full-length RPTPs.
No evidence for dimerization of full-length RPTPs
Subcloning of full-length RPTP coding sequences into the S2 cell expression vector pRmHa-3 had already been done for three of the RPTPs examined in this study. That left the full-length sequence of DLAR, which was cloned into the same vector by Neil Kreuger. Since the full-length constructs lack an epitope tag, antibodies were needed for each of the four RPTPs. mAbs against three of the four had previously been made in our lab. No antibody had been made against DLAR, so it was necessary to make one. We subcloned a portion of the DLAR extracellular domain including three FN III repeats into a baculovirus transfer vector. The vector was given to Peter Snow, who made the virus and expressed and purified the protein fragment. Susan Ou immunized mice with the polypeptide and performed the cell fusion. She supplied clones that were screened via ELISA. Finally, the strongest contenders were tested for their ability to IP DLAR. The resulting mouse mAb, 9D8, has proved efficient for IPs, Westerns and embryo staining.
The same experimental scheme used for cytoplasmic constructs (detailed above) was used with the full-length constructs. There was one procedural difference involving the method for Western blot development. The alkaline phosphatase method of detection that had been used previously was not sensitive enough for experiments with the full- length constructs. There appeared to be less protein on the blots, due to several possible causes. Perhaps the full-length constructs did not express as robustly as their cytoplasmic counterparts. The IPs may have been less efficient also; it may be more difficult to IP the larger full-length proteins (120-200 kD) than the smaller cytoplasmic versions (30- 80 kD). Finally, this size discrepancy may also have been an issue in the transfer to the membrane (efficient transfer may be more difficult with larger proteins). In any case, switching to the more sensitive ECL detection method made the bands easier to visualize.
This round of experiments was no more successful than the first in yielding evidence for RPTP dimerization. As before, cotransfections with all possible pairwise combinations of RPTPs were performed to assay for both hetero- and homodimers. The results of a typical experiment are shown in Figure 4. S2 cells were cotransfected with full-length DLAR and DPTP99A. Half the cell lysate was IPd with anti-DLAR and half with anti-DPTP99A. The anti-DLAR IP pellet was split in two, and one half was probed with anti-DLAR. The resulting DLAR band confirms expression. The other half was
probed with anti-DPTP99A, and the absence of a band suggests no DPTP99A co-IPd. To verify that DPTP99A expressed, one half of the DPTP99A pellet was probed with anti- DPTP99A. The resulting band, so strong it appears as a large smear, confirms robust expression. Probing the second half of the pellet with anti-DLAR reveals that no DLAR co-IPd.
Results from experiments with DPTP10D may have been misinterpreted as evidence in support of dimers were it not for the negative control. The apparent co-IPs that were observed in cases where DPTP10D was cotransfected with other RPTPs turned out to be an artifact of DPTP10D’s tendency to stick indiscriminately. In experiments where DPTP10D was cotransfected with a given RPTP “X,” an IP with anti-X followed by an anti-DPTP10D blot would show a DPTP10D band. By itself, this suggests a
DPTP10D/RPTP X dimer. The negative control was to IP a plate of cells transfected with only DPTP10D and IP with anti-X. If the dimer result is real, a DPTP10D band resulting from an anti-X IP should be seen only when X is coexpressed with DPTP10D. Because it appeared when DPTP10D was expressed alone, it must be attributable to cross-reactivity.
We observed that in addition to mAbs against all RPTPs, protein G+A beads alone were able to precipitate DPTP10D. This suggests that instead of cross-reactivity, the results were caused by a nonspecific interaction of DPTP10D with the G+A beads.
DISCUSSION
Crystallographic data suggests RPTPs form dimers
The first evidence to suggest RPTPs form dimers similar to their RTK counterparts comes from crystallographic studies. In each of two independent crystal structures, murine RPTPα exists as a symmetrical homodimer, where the helix-turn-helix (or
“wedge”) domain of one RPTP inserts into the active site of its dimer partner (Bilwes et al., 1996). The active site of both RPTPs is sterically blocked due to this interaction, theoretically rendering the enzymes inactive. Indeed, RPTPα studies have documented an inhibition of phosphatase activity upon dimerization (Jiang et al., 1999). RPTPs have a conserved wedge domain upstream of the first catalytic domain, denoted D1, suggesting
that downregulation of catalytic activity via dimerization-induced active site occlusion may represent a paradigm for regulation of RPTPs in general.
Subsequent studies have shown that dimer interactions can be complex, with a domain from one RPTP inserting into the catalytic domain of a different RPTP to form cross- species heterodimers (Blanchetot and den Hertog, 2000; Gross et al., 2002). For example, the second catalytic domain (D2) of PTPδ has been shown to bind to D1 of RPTPσ, and this results in an approximately 50% reduction in the catalytic activity of the latter (Wallace et al., 1998). Evidence also exists for formation of multimers, where strings of RPTPs are interconnected via interactions between wedge and active site domains of neighboring molecules (Iversen et al., 2002).
Intramolecular interactions also exist, as in the case of RPTP LAR. Although direct evidence for a LAR homodimer is lacking, the crystal structure suggests D1 of one LAR molecule may interact with D2 of a second (Nam et al., 1999). However, any model of dimer mediated regulation presupposes that all RPTPs form dimers in which the active site is obstructed, and crystal structures of other RPTPs indicate at least some do not. For example, RPTPµ does exist as a dimer in the crystal structure, but the wedge domain of one subunit of the dimer is not inserted into the catalytic cleft of the dyad-related monomer. Consequently, the active site remains in an open, uninhibited conformation (Hoffmann et al., 1997). Activity regulation through dimerization may thus be a feature of only a subset of RPTPs, and a goal of future work will be to develop a more
comprehensive account of which RPTPs belong in this category.
Evidence for in vivo dimer formation
Additional evidence for dimerization comes from recent cross-linking studies showing that RPTPα homodimerizes on the cell surface, suggesting that dimers form under normal physiological conditions (Jiang et al., 2000; Tertoolen et al., 2001). The same studies examined which of the RPTPα domains are necessary and/or sufficient for dimer formation. Remarkably, the results suggest no domain is absolutely required and that each is sufficient to mediate dimerization. Of all constructs tested, the ones where the wedge domain was included produced the most efficient dimer formation. Thus, although
multiple points of interaction are implicated in mediating dimerization of RPTPα, the wedge domain appears to play the major role.
Homodimers still form when the wedge domain is eliminated, albeit with the efficiency of dimerization significantly reduced. Thus, the region implicated by crystallographic studies as the major mediator of dimerization appears upon closer inspection to be important but not essential. Homodimers still form when the extracellular domain (ECD) is missing, so it too is not essential for dimerization.
However, it is sufficient for homodimerization, as determined with a fusion protein comprised of the ECD with a C-terminally fused GPI-linker for membrane insertion.
Finally, even the transmembrane domain when expressed by itself forms a homodimer.
The net result of these experiments suggests a “zipper” model wherein multiple interactions occur along length of the RPTP, each contributing to formation of a stable dimer.
Several lines of evidence support model of dimer mediated negative regulation of RTPT activity
Regulation of catalytic activity is well understood for the counterparts to the RPTPs, namely the RTKs. Following addition of ligand, RTKs dimerize and autophosphorylate, thereby switching on activity. A growing body of evidence suggests that RPTPs may work in the opposite manner, with dimerization inhibiting instead of promoting biological activity. Experiments with an EGF/CD45 chimera were the first to suggest this alternate theory of regulation. CD45 is an RPTP with no known ligand that is expressed on all nucleated hematopoietic cells and is required for TCR signaling in response to
engagement of antigen receptor. To study the effects of dimerization on CD45 activity, a chimeric protein was made with the extracellular and transmembrane domain of EGF fused to the intracellular domain of CD45. Addition of EGF to cells expressing the chimera results in a loss of TCR signaling, suggesting that induced dimerization extinguishes the catalytic activity of CD45 (Desai et al., 1993; Majeti et al., 1998).
The EGF/CD45 chimera data is consistent with results from another study involving induced dimerization. In this case, a point mutation in the extracellular domain of full- length RPTPα causes constitutive homodimerization via a disulfide bond. The FL-137C
mutant was transfected into fibroblasts from RPTPα knock-out mice to assay its ability to dephosphorylate (and thereby activate) its c-Src substrate. Results indicate the mutant activates c-Src significantly more weakly than WT RPTPα. Additional point mutations in the wedge region of FL-137C restore c-Src activation to a level comparable to that of WT.
Separately, the dimerization efficiency of RPTPα with the same wedge domain point mutations is shown to be significantly lower compared to WT RPTPα (Jiang et al., 1999).
In addition to biochemical evidence for a physiological role for dimerization, genetic data from our lab can be interpreted in a manner consistent with the dimer model.
RPTP DLAR is required for proper execution of several discrete axon guidance events during Drosophila nervous system development. One example is entry of the
intersegmental nerve b (ISNb) into the ventrolateral muscle field (VLM) after defasciculation from the intersegmental nerve (ISN). In embryos with a DLAR null mutation, the ISNb fails to enter the VLM approximately 25% of the time. However, in embryos that are doubly mutant for both DLAR and DPTP99A, this guidance error is reduced to about 2% (Desai et al., 1997). Other guidance mistakes observed in DLAR null mutants are not affected by removal of DPTP99A, suggesting a specificity in the suppression of this phenotype. This result implies that in the DLAR mutant, ISNb continues along its original trajectory instead of turning into the VLM because of inappropriate DPTP99A activity. Thus, the function of DLAR at this choice point is to counteract or suppress DPTP99A signaling. One way it could do that is by forming a heterodimer that, like other dimers studied, is catalytically inactive.
Taken together, these results support a theory of dimer mediated negative regulation of RPTP activity. In this model, RPTP activity would be switched on by ligands that dissociate dimers, or by intracellular processes such as phosphorylation that would transform the intracellular domains of existing dimers into an open conformation.
No evidence for stable dimer formation in S2 cells
Despite strong evidence that dimerization is a common means by which RPTP activity is modulated, we were unable to find any evidence for stable dimer formation under the conditions tested. If the RPTPs we studied are induced to dimerize by binding of ligands,
one possibility for these negative results is that these ligands are not expressed in S2 cells.
There are currently no known ligands for the four RPTPs examined in this study, so we were unable to assess whether ligand binding induces dimerization. A secreted factor has been identified that interacts with and inhibits the activity of RPTPβ (Meng et al., 2000), so it is reasonable to assume that ligands for other RPTPs do exist. The RPTPs’ large extracellular domains resemble those of cell adhesion molecules and have a high degree of variability, consistent with their having unique ligand binding specificities. Once a ligand for these RPTPs is identified and cloned, it can be added to the cell culture system to determine whether it facilitates dimer formation.
It is also possible that only a small fraction of the transiently expressed full-length RPTP makes it to the cell surface in our experiments. If most of the expressed protein is located intracellularly, it may be unable to assume the conformation necessary for engaging in dimer interactions.
Finally, our experimental conditions may permit transient dimer formation, but may not be conducive to formation of dimers stable enough to remain intact during the
immunoprecipitation process. It is also possible that a small fraction of dimers do remain intact but are below the limits of detection.
Figure 1. Sequence alignment of the N-terminal wedge of RPTPs’ membrane-proximal catalytic domains. Line one schematically illustrates the secondary structural elements observed in RPTPαD1. Line two gives the consensus sequence conserved across the large family of D1 domains. The numbering scheme refers to the short variant of murine RPTPα with the first residue of the signal peptide numbered as 1. The stars indicate residues involved in active-site-directed dimeric interactions. Residues boxed in black are common to at least five aligned sequences. (m is mouse, h is human, r is rat, and d is Drosophila). The first 14 homologous sequences contrast sharply with the lack of
sequence conservation in the final segments. The most distinguishing feature of the RPTP N-terminal wedge involves a two-amino-acid insertion (Asp 227 and Asp 228) into the tip of the non-receptor-like PTP N-terminal segment. The second to the last line corresponds to the numbering scheme of human PTP1B. The final line emphasizes the structural elements observed in the PTP1B crystal structure.
Source: Nature (1996) 382, 559
Figure 2. psmyc schematic
Figure 3. DLAR homodimer experiment
α-myc blot
IP:
M=α-myc R=α-rho
DLARrm yc
+ DLARrho
DLARmyc + DLARrho DLARmyc
DLARmyc
DLARrho DLARrho
M R M R M M R R α-rho blot DLARrm
yc + DLARrho
DLARmyc + DLARrho
Figure 4. DLAR/DPTP99A heterodimer experiment
Dlar 99A Dlar 99A
α-DLAR blot α-DPTP99A blot
IP—α:
DPTP99A + DLAR DLAR
DPTP99A
DPTP99A + DLAR DLAR
DPTP99A
DPTP99A +DLAR DLAR
DPTP99A
DPTP99A + DLAR DLAR
DPTP99A
Table 1. S2 cell transfection experiments
Heterodimer experiments Homodimer experiments
DPTP69D FL + DPTP10D cyto myc DPTP69D FL + DPTP69D cyto myc DPTP69D FL + DLAR FL myc DPTP99A FL + DPTP99A cyto myc DPTP69D FL + DLAR cyto myc DPTP10D FL + DPTP10D cyto myc DPTP69D FL + DPTP99A FL DLAR FL + DLAR cyto myc
DPTP69D FL + DPTP10D FL DLAR cyto rho + DLAR cyto myc DPTP99A FL + DPTP69D cyto myc
DPTP99A FL + DLAR FL myc DPTP99A FL + DLAR cyto myc DPTP10D FL + DLAR cyto myc
cyto=cytoplasmic domain, FL=full-length
Chapter 3
Biochemical analysis of potential interactions between RPTPs and Robo receptors
Abstract
Robo is the guidance receptor for the Drosophila midline repellent Slit. Robo signaling is required to keep axons from inappropriately crossing the midline. Robo and receptor- linked tyrosine phosphatases (RPTPs) are expressed in growth cones during axonogenesis and several genetic observations suggest that Robo signaling involves the activity of RPTPs DPTP10D and/or DPTP69D. We show here that Robo is phosphorylated on one or more tyrosine residues, a feature characteristic of phosphatase substrates. No evidence was seen, however, for a biochemical interaction between Robos and “substrate trap”
versions of the RPTPs, leaving us to conclude that RPTPs’ contribution to Robo signaling may be downstream of Robo.
INTRODUCTION
Many extracellular signals, once transduced to the cell interior, are ultimately manifest as a change in the tyrosine phosphorylation status of specific regulatory proteins. Two classes of enzymes, protein tyrosine kinases (PTKs) and protein tyrosine phosphatases (PTPs), work coordinately to regulate intracellular phosphotyrosine levels. Together they are involved in a diverse number of cellular processes including differentiation,
proliferation and axon guidance. PTPs can be divided into two classes: receptor-like PTPs that span a membrane, and cytoplasmic PTPs. The receptor PTPs (RPTPs) have extracellular domains comprised of N-terminal immunoglobulin (Ig) domains and membrane-proximal fibronectin type III (FNIII) repeats. The structure of their ectodomain is thus similar to that of cell adhesion molecules, implicating them in participating in cell-cell, cell-extracellular matrix adhesive events underlying axon pathfinding. The intracellular domain of most RPTPs is comprised of two protein
tyrosine phosphatase catalytic domains denoted D1 and D2. The membrane-proximal D1 domain has most or all of the catalytic activity, while the C-terminal D2 has little or none.
The high degree of sequence conservation in D2 among members of the RPTP family suggests it may have an indispensable function unrelated to catalysis. Among the proposed roles for D2 are binding of substrates and/or downstream regulatory factors.
This domain may also regulate activity by inserting into the active site of dimerized RPTPs, thereby precluding substrates from binding. Recent findings that RPTPα homodimerizes in vivo are consistent with this model (Tertoolen et al., 2001), yet we have found no evidence for dimerization in the RPTPs controlling axon guidance in Drosophila (see Chapter 2).
The role of RPTPs in axon guidance and synaptogenesis has been studied extensively in Drosophila. During axonogenesis, four RPTPs (DPTP69D, DPTP99A, DPTP10D and DLAR) are selectively expressed on CNS axons and growth cones (Tian et al., 1991). They interact with each other in various well-defined ways to facilitate a given axon’s navigation from one choice point to the next en route to a final target destination. In some cases they work antagonistically, while in others they collaborate to regulate guidance decisions (Desai et al., 1997; Sun et al., 2001). Mutations in these
enzymes result in characteristic axon guidance phenotypes. However, a mutation in any single Rptp gene usually has no effect on guidance decisions. Most axon guidance
phenotypes are observed only under conditions where two or more Rptps are mutant. This suggests partial redundancy in the function of these genes.
Ptp69D and Ptp10D provide one example of the redundant nature of the Rptp genes.
When either gene is mutated individually, no phenotype is observed and embryos are both viable and fertile. However, in embryos doubly mutant for both Ptp69D and Ptp10D, a strong phenotype in the ventral nerve cord is observed. The nerve cord is the fly
equivalent of the vertebrate spinal cord and is comprised of a ladder-like scaffold of axons. The upright parts of the ladder are formed by longitudinal axon bundles called connectives, while the rungs are formed by segmentally reiterated pairs of bundles called commissures. The decision to cross the nerve cord midline underlies the formation of these two bundle types: axons that choose to remain on one side of the midline form longitudinals, while those that make the decision to cross form commissures. In the Ptp69D Ptp10D double mutant, axons that normally do not cross the midline follow abnormal midline-crossing pathways, resulting in a highly disorganized nerve cord (Sun et al., 2000).
The Ptp69D Ptp10D double mutant phenotype resembles that of embryos mutant for Roundabout (Robo). Robo is the receptor for a protein called Slit, which is secreted by midline glial cells (Brose et al., 1999). Slit, a midline repellent, mediates the repulsion of Robo-expressing growth cones (Kidd et al., 1999). Robo is expressed in all nerve cord axons, so how is it that some are able to cross the midline? The answer has to do with a third protein called Commissureless (Comm) which is expressed in commissural axons and midline glial cells. Recent findings show that the function of Comm is to sort Robo to endosomes, where it is degraded (Keleman et al., 2002; Myat et al., 2002). As a result of this sorting function, no Robo makes it to the cell surface in Comm-expressing axons.
This makes them impervious to the repellent qualities of Slit, enabling them to cross the midline. Comm expression ceases after the axons have crossed, and the axons are once again sensitive to Slit due to Robo on their cell surface. This keeps them from
inappropriately recrossing the midline.
Several lines of genetic evidence show that DPTP69D and/or DPTP10D may be positive regulators of Robo signaling (Sun et al., 2000). First, a Ptp69D Ptp10D double mutant has a midline crossing phenotype that resembles that of a robo mutant in some respects. Second, the same double mutation partially suppresses the phenotype of comm mutants. In comm- embryos, all Robo localizes to the cell surface because there is no Comm sorting it to endosomes. Consequently, all axons are sensitive to Slit and repelled from the midline. Introduction of the Rptp double mutant into this background results in some axons crossing the midline, suggesting that Robo’s signaling has been
compromised. In the third line of evidence, the double mutant is introduced into a slit- background. When one copy of slit is removed, enough repulsive signaling remains to keep axons from crossing the midline inappropriately. That is no longer true when Rptps are removed as well, and less efficient Robo signaling is a plausible interpretation.
No substrates have yet been identified for the RPTPs, but the above genetic data led us to speculate that Robo may be a substrate for DPTP69D and/or DPTP10D. Consistent with this, Robo has an intracellular phosphotyrosine consensus motif. If this prediction is correct, robust signaling through Robo may require not only Slit binding but also
dephosphorylation by an RPTP. Lending credence to this theory, Robo appears to be negatively regulated by Abl tyrosine kinase signaling, and a Y-to-F mutation in this motif produces a gain-of-function phenotype (Bashaw et al., 2001). Additional features of Robo that make it suitable for consideration as an RPTP substrate include its being expressed at the same time (axonogenesis) and in the same place (growth cones) as RPTPs.
Furthermore, as a surface protein it has the same subcellular localization as RPTPs.
A prerequisite for any RPTP substrate is that it is tyrosine phosphorylated. We show here that Robo1 and Robo2 are phosphorylated by the tyrosine kinases Src and Abl.
This finding encouraged us to pursue experiments to test whether Robos are RPTP
substrates. We cotransfected S2 cells with Robo and a substrate trap version of DPTP10D (DPTP10D trap) or DPTP69D (DA3). The substrate traps have a point mutation that causes them to remain bound to substrate (Flint et al., 1997). In some cases trap proteins are rendered useless because phosphorylation on certain residues sterically hinders the binding of substrate. Our results show that RPTP traps used in this study are not phosphorylated, and thus not precluded from binding substrate. In spite of this, and
contrary to data favoring an enzyme/substrate relationship, our results show no evidence that Robo is a substrate for RPTPs. It is possible this is due to a requisite co-factor being absent from the cell culture system. However, it is equally plausible that RPTPs
contribute to Robo signaling by dephosphorylating other proteins in the Robo pathway.
RESULTS
The approach we took to determine whether Robo is an RPTP substrate was as follows.
First, ascertain whether Robos are phosphorylated on tyrosine residues. If they are, they meet a basic prerequisite for potential RPTP substrates. Provided this result is positive, cotransfect Robo and RPTP substrate trap constructs, immunoprecipitate the mutant RPTPs and look for Robo. If Robo co-IPs, it suggests there is an interaction between the proteins. That would be the basis for proceeding with the next set of experiments,
designed to determine the particular type of interaction. Specifically, we are interested in looking for evidence of an enzyme-substrate interaction. This involves two separate cotransfections, one with wild type RPTP and Robo, the other with substrate trap RPTP and Robo. Once again, the RPTPs are immunoprecipitated and the associated Robo is visualized on a Western blot. If the amount of Robo that co-IPs with the RPTP trap is significantly greater than what copurifies with WT RPTP, this would be evidence in support of the enzyme-substrate interaction. However, if the amount of copurifying Robo is the same in both cases, it would suggest a different type of interaction.
As experiments testing for Robo phosphorylation began, the RPTP trap plasmids were constructed. The nature of the trap mutation can be understood in the context of the dephosphorylation reaction. All phosphatases have a conserved catalytic domain of approximately 240 residues that is characterized by a signature motif:
(I/V)HCXAGXXR(S/T)G. This motif contains an invariant cysteine residue that is essential for catalytic activity. It acts as a nucleophile and attacks a phosphorus-oxygen bond in the substrate, leading to formation of a thiol-phosphate intermediate. Binding of substrate induces a conformational change in the enzyme, causing movement of a loop that forms one side of the active site cleft. This results in a more closed structure around
